In the spirit of "horses for courses", is the SE software the best platform for the meta site? At MathOverflow, they use a traditional forum for the meta site and I feel that it works very well. Things on meta don't always coalesce into straightfoward "question and answer"; sometimes a discussion is warranted, sometimes a post is simply for information - as an example, I just posted a proposed tag merge; on MO there is a permanent thread where people can post such proposals meaning that it doesn't bother anyone else and the moderators have a single place to check.
What does anyone else think?
I'd like to add a couple of points in favour of a traditional forum:
The purpose of the meta site is to deal with issues arising from the main site. From experience at MathOverflow (the most successful of the first round SE sites), these are often issues where there is some disagreement and where it is important to be sure that different viewpoints are heard. So discussions are not just encouraged, they are necessary.
The comparisons with how it works on SO are invalid. The SO trilogy is not community led in the same way as this site is community led. The SO trilogy has a (benevolent) dictatorship and so the dynamics of SO are different to the dynamics here. On SO it's much more "What's the rules on this?" or "How do I do that?" whereas here it's going to be "What should the rule on this be?" or "Should we ask SE for the ability to do that?". For example, after spending yesterday on tex and meta-tex, I have no clear idea of who's "in charge" or even if anyone is "in charge". MathOverflow [MO] also has a benevolent dictatorship, but many of the early users "out-ranked" said dictator in real life and so felt no compunction about speaking their mind! So, I feel, the comparison with MO is much stronger than with SO and having a traditional forum for meta has really helped us shape MO to the useful tool that it is for mathematicians.
The canard about forums being a disaster area is also not relevant. My experience of forums (fora?) is entirely positive because (I think) those forums exist for something outside of themselves. The nForum (which I run) exists to serve the purpose of the nLab, and so it's easy to identify inappropriate behaviour. Similarly meta.MO exists to serve MO and again it's easy to identify inappropriate behaviour. A meta.TeX forum would have the same link. No sane person would "hang around" on meta.TeX just to troll it. Even the insane ones (and we get plenty of those in mathematics) tend to keep away.
I once saw a picture on Jeff Attwood's blog where he drew a Venn diagram with wikis, forums, and blogs as three big circles whose intersection was SO. I completely agree with that diagram, but I draw a completely different conclusion from it. The SO software combines features of all three, but also loses features of each of them. So if you want the features that SO provides, it's fantastic. But as soon as you need something a little more like one of the other three types, the SO software creaks at the seams.
But that's not a problem. As a die-hard Linux user, that's just how the best systems work! Lots of small programs, each doing one thing and doing it well, add up to a global system that doesn't just beat the competition, it beats it, whips it, and puts it in the oven to make a delicious cake out of the pieces.